A fast monocular visual odometry pose estimation method for self-driving vehicles

Abstract

Aiming at the problem that the traditional pose estimation algorithm of monocular odometry cannot meet the needs of real-time localization of self-driving vehicles, a fast pose estimation algorithm based on the improvement of the fundamental matrix is proposed. By optimizing the calculation process of the fundamental matrix, the real-time performance of the algorithm is improved. The fundamental matrix with 8 unknown parameters is first obtained, and then the feature matching point pairs are used to solve the fundamental matrix. Through simulation experiments, the efficiency and accuracy of the algorithm are analyzed, and then it is compared with the existing algorithms. Experimental results show that the proposed algorithm can improve the speed of motion estimation by 4 times without the reduction of the accuracy of motion estimation. The study provides certain reference value to the real time application of the visual odometry of self-driving vehicles.

Keywords

• vehicle engineering
• visual odometry
• pose estimation
• 5 points algorithm
• fundamental matrix